Classification on PASCAL VOC 2007 (test)
93.1mAP (%)MELM
Evaluation results
| Method | Links | ||||||||||||||||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| MELMCNN=VGG162019.02 | 93.1 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| WCCNBackbone=VGG16, Stages=3-stage2016.11 | 90.9 | 94.2 | 94.8 | 92.8 | 91.7 | 84.1 | 93 | 93.5 | 93.9 | 80.7 | 91.9 | 85.3 | 97.5 | 93.4 | 92.6 | 96.1 | 84.2 | 91.1 | 83.3 | 95.5 | 89.6 | — | — | — | — | — | — | — | |
| WCCNCNN=VGG162019.02 | 90.9 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| WCCNBackbone=VGG16, Stages=2-stage2016.11 | 89.9 | 93.4 | 93.7 | 92 | 91 | 83.1 | 91.5 | 92.7 | 93.5 | 79.3 | 90.7 | 83.1 | 96.9 | 92.9 | 91.2 | 95.9 | 82.4 | 90.3 | 81.3 | 95.1 | 88.3 | — | — | — | — | — | — | — | |
| WSDDNModel Size=L2015.11 | 89.7 | 93.3 | 93.9 | 91.6 | 90.8 | 82.5 | 91.4 | 92.9 | 93 | 78.1 | 90.5 | 82.3 | 95.4 | 92.7 | 92.4 | 95.1 | 83.4 | 90.5 | 80.1 | 94.5 | 89.6 | — | — | — | — | — | — | — | |
| WSDDNCNN=VGG162019.02 | 89.7 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| VGG16-net2016.11 | 89.3 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| VGG-VD162015.11 | 89.3 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| VGG16CNN=VGG162019.02 | 89.3 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| WSDDN2016.11 | 89 | 95 | 92.6 | 91.2 | 90.4 | 79 | 89.2 | 92.8 | 92.4 | 78.5 | 90.5 | 80.4 | 95.1 | 91.6 | 92.5 | 94.7 | 82.2 | 89.9 | 80.3 | 93.1 | 89.1 | — | — | — | — | — | — | — | |
| WSDDNModel Size=Ensemble2015.11 | 89 | 95 | 92.6 | 91.2 | 90.4 | 79 | 89.2 | 92.8 | 92.4 | 78.5 | 90.5 | 80.4 | 95.1 | 91.6 | 92.5 | 94.7 | 82.2 | 89.9 | 80.3 | 93.1 | 89.1 | — | — | — | — | — | — | — | |
| TripletTraining Epochs=980, Backbone=ResNet-50, Evaluation Protocol=Linear SVM2023.10 | 88.3 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | 43.6 | 56.2 | 64.6 | 73.8 | 79.6 | — | — | |
| SCEK (images per class)=full, Pre-training epochs=2002022.12 | 88.2 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| DINOTraining Epochs=800, multi-crop=true, Backbone=ResNet-50, Evaluation Protocol=Linear SVM2023.10 | 88.2 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | 45.6 | 58.4 | 66.6 | 74.8 | 79.6 | — | — | |
| CARPTraining Epochs=400, multi-crop=true, Backbone=ResNet-50, Evaluation Protocol=Linear SVM2023.10 | 88.2 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | 47.1 | 59.8 | 67.3 | 75.8 | 80 | — | — | |
| CARPTraining Epochs=200, multi-crop=true, Backbone=ResNet-50, Evaluation Protocol=Linear SVM2023.10 | 88 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | 46 | 58.3 | 66.5 | 75.5 | 79.5 | — | — | |
| WCCNBackbone=AlexNet, Stages=3-stage2016.11 | 87.8 | 93.1 | 91.1 | 89.6 | 88.9 | 81 | 89.6 | 90.7 | 91.2 | 76.4 | 89.2 | 80.8 | 92.2 | 90.1 | 89 | 92.7 | 82 | 89.3 | 78.1 | 92.8 | 89.1 | — | — | — | — | — | — | — | |
| WCCNCNN=VGGF/AlexNet2019.02 | 87.8 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MELMCNN=VGGF/AlexNet2019.02 | 87.8 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| WCLK (images per class)=full, Pre-training epochs=2002022.12 | 87.8 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SwAVK (images per class)=full, Pre-training epochs=2002022.12 | 87.5 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MoCo-v3Training Epochs=1000, Backbone=ResNet-50, Evaluation Protocol=Linear SVM2023.10 | 87.4 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | 46.6 | 59.6 | 67 | 75.4 | 80.2 | — | — | |
| SupervisedEpochs=902021.07 | 87.27 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | 82.26 | 84 | 85.13 | |
| WCCNBackbone=AlexNet, Stages=2-stage2016.11 | 87.1 | 92.8 | 90.3 | 89.3 | 88.2 | 80.4 | 89.4 | 90 | 90.4 | 75.3 | 88.1 | 80.1 | 91.3 | 89.1 | 88.3 | 91.2 | 80.6 | 88.5 | 77.8 | 92.2 | 88.7 | — | — | — | — | — | — | — | |
| WSDDNModel Size=M2015.11 | 86.4 | 93.9 | 91 | 90.4 | 89.3 | 72.7 | 86.4 | 91.9 | 91.5 | 73.8 | 85.6 | 74.9 | 91.9 | 91.5 | 89.9 | 94.5 | 78.6 | 85 | 78.6 | 91.5 | 85.7 | — | — | — | — | — | — | — | |
| ReSSLEpochs=2002021.07 | 86.31 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | 79.17 | 81.96 | 83.81 | |
| ReSSLK (images per class)=full, Pre-training epochs=2002022.12 | 86.3 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SCEK (images per class)=64, Pre-training epochs=2002022.12 | 85.5 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| PCL V2Epochs=2002021.07 | 85.43 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | 78.34 | 80.72 | 82.67 | |
| PCLv2K (images per class)=full, Pre-training epochs=2002022.12 | 85.4 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| PCL v2Training Epochs=200, Backbone=ResNet-50, Evaluation Protocol=Linear SVM2023.10 | 85.4 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | 47.9 | 59.6 | 66.2 | 74.5 | 78.3 | — | — | |
| WSDDNModel Size=S2015.11 | 85.3 | 92.5 | 89.9 | 89.5 | 88.3 | 66.5 | 83.6 | 92.1 | 90.3 | 73 | 85.7 | 72.6 | 91.4 | 90.1 | 89 | 94.4 | 78.1 | 86 | 76.1 | 91.1 | 85.5 | — | — | — | — | — | — | — | |
| WSDDNCNN=VGGF/AlexNet2019.02 | 85.3 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| WCLK (images per class)=64, Pre-training epochs=2002022.12 | 85 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MoCo V2Epochs=2002021.07 | 84.6 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | 76.14 | 79.16 | 81.52 | |
| MoCov2K (images per class)=full, Pre-training epochs=2002022.12 | 84.6 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SwAVK (images per class)=64, Pre-training epochs=2002022.12 | 84.4 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| ReSSLK (images per class)=64, Pre-training epochs=2002022.12 | 83.8 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SCEK (images per class)=32, Pre-training epochs=2002022.12 | 83.1 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| WCLK (images per class)=32, Pre-training epochs=2002022.12 | 83 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| PCLv2K (images per class)=64, Pre-training epochs=2002022.12 | 82.7 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SPP-net2014.06 | 82.44 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Chatfield et al.2014.06 | 82.42 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SPPnet2016.11 | 82.4 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Alexnet2016.11 | 82.4 | 95.3 | 90.4 | 92.5 | 89.6 | 54.4 | 81.9 | 91.5 | 91.9 | 64.1 | 76.3 | 74.9 | 89.7 | 92.2 | 86.9 | 95.2 | 60.7 | 82.9 | 68 | 95.5 | 74.4 | — | — | — | — | — | — | — | |
| SPP2015.11 | 82.4 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| VGG-S2015.11 | 82.4 | 95.3 | 90.4 | 92.5 | 89.6 | 54.4 | 81.9 | 91.5 | 91.9 | 64.1 | 76.3 | 74.9 | 89.7 | 92.2 | 86.9 | 95.2 | 60.7 | 82.9 | 68 | 95.5 | 74.4 | — | — | — | — | — | — | — | |
| AlexNetCNN=VGGF/AlexNet2019.02 | 82.4 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| ReSSLK (images per class)=32, Pre-training epochs=2002022.12 | 82 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SwAVK (images per class)=32, Pre-training epochs=2002022.12 | 81.9 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MoCov2K (images per class)=64, Pre-training epochs=2002022.12 | 81.5 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| PCLv2K (images per class)=32, Pre-training epochs=2002022.12 | 80.7 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| WCLK (images per class)=16, Pre-training epochs=2002022.12 | 80.2 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| ImageNetfine-tune layers=all2016.03 | 79.9 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| VGG-M-10242015.11 | 79.9 | 91.4 | 86.9 | 89.3 | 85.8 | 53.3 | 79.8 | 87.8 | 88.6 | 59 | 77.2 | 73.1 | 85.9 | 88.3 | 83.5 | 91.8 | 59.9 | 81.4 | 68.3 | 93 | 74.1 | — | — | — | — | — | — | — | |
| ImageNet labelsfrozen layers=none, fine-tuned layers=all2016.11 | 79.9 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Supervisedbackbone=AlexNet, fine-tuning=true2017.08 | 79.9 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| ImageNet labelsTrained layers=all2018.03 | 79.9 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| ImageNet labelsFinetuning layers=all, Backbone=AlexNet2019.11 | 79.9 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SCEK (images per class)=16, Pre-training epochs=2002022.12 | 79.5 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MoCov2K (images per class)=32, Pre-training epochs=2002022.12 | 79.2 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| ReSSLK (images per class)=16, Pre-training epochs=2002022.12 | 79.2 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| ImageNetfine-tune layers=fc6-82016.03 | 78.9 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| ImageNet labelsfrozen layers=conv5, fine-tuned layers=fc6-82016.11 | 78.9 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| ImageNet labelsTrained layers=fc6-82018.03 | 78.9 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| ImageNet labelsFinetuning layers=fc6-8, Backbone=AlexNet2019.11 | 78.9 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| PIRLPre-training Dataset=YFCC1M, Backbone=ResNet-50, Evaluation Protocol=Linear classifier on fixed representations2019.12 | 78.8 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SwAVK (images per class)=16, Pre-training epochs=2002022.12 | 78.4 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| PCLv2K (images per class)=16, Pre-training epochs=2002022.12 | 78.3 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Krizhevsky et al.Pretraining time=3 days, Source=ImageNet, Supervision=labeled classes2017.08 | 78.2 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Oquab et al.2014.06 | 77.7 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Oquab et al.2016.11 | 77.7 | 88.5 | 81.5 | 87.9 | 82 | 47.5 | 75.5 | 90.1 | 87.2 | 61.6 | 75.7 | 67.3 | 85.5 | 83.5 | 80 | 95.6 | 60.8 | 76.8 | 58 | 90.4 | 77.9 | — | — | — | — | — | — | — | |
| Oquab et al.2015.11 | 77.7 | 88.5 | 81.5 | 87.9 | 82 | 47.5 | 75.5 | 90.1 | 87.2 | 61.6 | 75.7 | 67.3 | 85.5 | 83.5 | 80 | 95.6 | 60.8 | 76.8 | 58 | 90.4 | 77.9 | — | — | — | — | — | — | — | |
| VGG-F2015.11 | 77.4 | 88.7 | 83.9 | 87 | 84.7 | 46.9 | 77.5 | 86.3 | 85.4 | 58.6 | 71 | 72.6 | 82 | 87.9 | 80.7 | 91.8 | 58.5 | 77.4 | 66.3 | 89.1 | 71.3 | — | — | — | — | — | — | — | |
| SeLaFinetuning layers=all, Backbone=Larger AlexNet variant, Clustering config=3k x 10, Auxiliary task=Rotation2019.11 | 77.2 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| ImageNetfine-tune layers=fc82016.03 | 76.8 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MoCov2K (images per class)=16, Pre-training epochs=2002022.12 | 76.1 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Zeiler & Fergusreported_by=[27]2014.06 | 75.9 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SeLaFinetuning layers=all, Backbone=Larger AlexNet variant, Clustering config=3k x 10, Variant=¯2019.11 | 75.9 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SeLaFinetuning layers=fc6-8, Backbone=Larger AlexNet variant, Clustering config=3k x 10, Auxiliary task=Rotation2019.11 | 75.6 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SeLaFinetuning layers=all, Backbone=Larger AlexNet variant, Clustering config=3k x 102019.11 | 75.3 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| RotNet+retrievalFinetuning layers=all, Backbone=Larger AlexNet variant2019.11 | 74.7 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SeLaFinetuning layers=fc6-8, Backbone=Larger AlexNet variant, Clustering config=3k x 10, Variant=¯2019.11 | 74.4 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| DeepClusterFinetuning layers=all, Backbone=AlexNet2019.11 | 73.4 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SeLaFinetuning layers=fc6-8, Backbone=Larger AlexNet variant, Clustering config=3k x 102019.11 | 73.1 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| DeeperClusterPre-training Dataset=YFCC100M, Backbone=VGG-16, Evaluation Protocol=Linear classifier on fixed representations2019.12 | 73 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| RotNetFinetuning layers=all, Backbone=AlexNet2019.11 | 73 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| RotNetTrained layers=all2018.03 | 72.97 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| CC+VGGFinetuning layers=all, Backbone=AlexNet2019.11 | 72.5 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| RotNet+retrievalFinetuning layers=fc6-8, Backbone=Larger AlexNet variant2019.11 | 72.5 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MILinearCNN=VGGF/AlexNet2019.02 | 72 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| DeepClusterFinetuning layers=fc6-8, Backbone=AlexNet2019.11 | 72 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| JigsawPre-training Dataset=YFCC100M, Backbone=ResNet-50, Evaluation Protocol=Linear classifier on fixed representations2019.12 | 71 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| RotNetFinetuning layers=fc6-8, Backbone=AlexNet2019.11 | 70.9 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| RotNetTrained layers=fc6-82018.03 | 70.87 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Context 2Finetuning layers=all, Backbone=AlexNet2019.11 | 69.6 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Norrozi et al.Pretraining time=2.5 days, Source=ImageNet, Supervision=context2017.08 | 68.6 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Countingbackbone=AlexNet, fine-tuning=true2017.08 | 67.7 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| CountingTrained layers=all2018.03 | 67.7 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Jigsawbackbone=AlexNet, fine-tuning=true2017.08 | 67.6 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — |